ESTRO 2023 - Abstract Book

S548

Sunday 14 May 2023

ESTRO 2023

Results 92 patients (57 brain metastases and 35 primary brain tumors) with a median age of 69 (range: 32-86) were enrolled. The median KPS was 90, 62% were female, 65% were Hispanic, and 55% had a high school or equivalent level of education. Of those enrolled, 41% successfully completed neurocognitive evaluation in Spanish. At 3 months, there was a statistically significant change in HVLT-R-delayed recall (DR) (Z score: 3.97, p<0.001), HVLT-R-immediate recall (IMM) (3.68, p<0.001), and TMT A (-4.21, p<0.001). At 6 months, there was a statistically significant decline in SDMT (-5.45, p=0.001) across all patients; on subset analysis, this was driven by a significant decline in brain metastasis patients only (-6.31, p=0.003). Among 35 patients who completed the patient feedback survey, the majority found the system and questions easy to understand (100%), easy to use (95%), and relevant to their care (71%). Most patients reported that neurocognitive evaluation improved discussions with clinicians (62%), made them feel more in control of their own care (67%), and 81% reported that they would recommend the app to other patients with CNS malignancies. Conclusion Implementation of this novel interactive, multi-language, app-based neurocognitive evaluation program demonstrated changes post-radiotherapy in a majority of patients across multiple domains, warranting further dosimetric and treatment technique evaluation. This app-based assessment was easy to use, perceived as useful to a patient’s care, and highly recommended to other patients. PD-0657 Sub-second speed 4D-CT image registration using deep learning T. van der Meulen 1 , O. Pastor-Serrano 1 , S. Habraken 2 , Z. Perkó 1 1 Delft University of Technology, Radiation Science and Technology, Delft, The Netherlands; 2 Erasmus University Medical Center, Department of Radiotherapy, Rotterdam, The Netherlands Purpose or Objective From contour propagation to dose accumulation, fast image registration is crucial for treating moving targets. Traditional registration methods optimize an objective function per pair of images, typically resulting in minutes of computation times, making them inapplicable in online and real-time adaptive treatment schemes. To overcome this challenge, our work presents a deep learning based accurate and robust registration model that can register pairs of images in few tens of milliseconds. Materials and Methods We use a Laplacian Pyramid Image Registration Network (LaPy) for image deformation and contour propagation between breathing phases of 4DCT scans from non-small cell lung cancer patients. The network takes a fixed and moving image as input and predicts the full deformation field warping the moving image to match the fixed one. The LaPy model is trained using 65 planning and repeat 4D-CT scans from 20 different patients. Each 4D-CT is discretized into 10 phases per breathing cycle, resulting in 100 image pairs (90 combinations between phases plus 10 identity registrations). During training, all images are normalized to the range [0,1] and shifted ±10 mm for data augmentation. For model evaluation, 7 additional 4D-CTs unseen during training from 7 patients from the same dataset are used to compare the LaPy model to the registration software Elastix (computing b-spline deformations with parameters from a previously published study). We use the mean squared error (MSE) between the fixed and warped normalized image values to measure image similarity, and the mean surface distance (MSD) and Dice score to assess contour overlap for the tumor, lungs, heart, carina and esophagus. To test model robustness, we additionally evaluate the MSE on 2 datasets of 7 and 5 patients with 4D-CTs recorded with a different machine. Results The LaPy model can accurately infer the deformation field between the 4D phases for the 7 testing patients, achieving very low MSE of 6.029E-5±6.065E-5, comparable to Elastix’s error of 6.026E-5±3.014E-5. For the 2 external datasets from a different machine and location, the MSE is similarly low, 6.96E-5±3.95E-5 and 7.02E-5±3.80E-5, respectively. Table 1 shows that the model can reach significantly lower MSD and higher Dice score compared to the input moving images, demonstrating LaPy’s ability to model breathing anatomical deformations. Figure 1 further underlines the excellent agreement between the predicted and fixed images. With average GPU prediction times of 23±2 ms, the LaPy model significantly outperforms Elastix’s computing times of 248±11 s. Poster Discussion: Imaging